Simple Semantic-Aided Few-Shot Learning

1. WordNet
 

import nltk

def get_definition(word):
    # 获取与给定单词相关的同义词集合
    synsets = wn.synsets(word)
    
    # 如果没有找到同义词集合，返回提示信息
    if not synsets:
        return f"No definition found for '{word}'."
    
    # 输出每个同义词集合的名称和定义
    definitions = []
    for syn in synsets:
        definitions.append(f"{syn.name()}: {syn.definition()}")
    
    return definitions

# 示例
word = 'Robin'
definitions = get_definition(word)

for definition in definitions:
    print(definition)

结果： 

robin.n.01: small Old World songbird with a reddish breast
robin.n.02: large American thrush having a rust-red breast and abdomen

      小型的旧世界鸣鸟，胸部呈红色。

     大型美国阔嘴鸟，胸部和腹部呈铁锈红色。

2. 实验结果

5way-1shot:  ResNet

        FC-100:

      CIFAR-FS:

     miniimagenet:

GPT-3:

zero-shot: GPT-4